Research On Tarck-Before-Detect Of Weak Targets Algorithm Based On Improved Particle Filter

With the gradual development and improvement of the target s stealth technology,the importance of detecting and tracking weak targets in the fields of modern military such as coordinated operations and navigation guidance has gradually emerged.When the signal energy radiated by the targets on the shooting range is weak or the environment is complicated,which leads to the signal noise ratio of the target accepted by the radar station is low.For this problem,the conventional Detect-Before-Track(DBT)method cannot effectively solve the detection and tracking of weak targets.However,the corresponding Track-Before-Detect(TBD)method can achieve weak target detection and tracking.The method of TBD combines the detection and tracking,and does not filtering on the single-frame observation data,but accumulates energy along a path which has larger energy.Particle filter(Particle filtering,PF)in solving nonlinear and non-Gaussian has a better tracking precision and detection performance.Therefore,the thesis focuses on the research of weak target detection and tracking based on improved PF algorithm.1)The thesis first introduces the TAD and TBD methods,and elaborates on the respective characteristics of the two methods in solving the target detection and tracking,focusing on the advantages of TBD in dealing with weak tar get detection and tracking.Secondly,the basic process of Bayesian recursive filtering based on Bayesian estimation theory framework is introduced,and the problems in the implementation process are analyzed.Then,based on the sequential importance sampling method,this method can effectively solve the problem that the analytical solution cannot be obtained in the Bayesian recursive filtering process.The influence of the selection of the importance probability density function on the state estimation performance is analyzed.Furthermore,the introduction of the resampling method can effectively overcome the particle degradation phenomenon.Finally,the standard PF implementation process is given,and the advantages of PF algorithm in solving nonlinear and non-Gaussian problems are verified by experimental simulation.2)This thesis discusses and analyzes the problems of tracking before detection of maneuvering weak targets.Firstly,described the motion model and measurement model of the multi-model particle filter track before detection(MMPF-TBD)algorithm.Secondly,this thesis proposes an improved multi-model particle filter track before detection(IMMPF-TBD)based on MMPF-TBD algorithm.The IMMPF-TBD algorithm obtains particle acceleration information by sampling particles in a mixed state,and then establishes a state transition matrix for each particle according to the characteristics of the particles.Further maneuvering models are derived to quickly match the maneuvering motion of the target.Finally,the experimental simulation is used to test the detection effect of the algorithm on weak targets.3)In the process of implementing PF algorithm,the resampling process copies particles with large weights and ignores particles with small weights,which constantly leads to the problem of particle diversity reduction.Therefore,the standard particle filter track before detection(PF-TBD)algorithm will affect the tracking effect and detection performance of weak targets.For this problem,an optimized particle filter track before detection(OPF-TBD)algorithm is proposed.The core of this algorithm is to introduce the observation information of the target state by using the idea of Extended Kalman Filter(EKF)in the process of sampling the particles,and cross-mutation operation on the particles with smaller weights,so that the sampled The particle distribution can further approximate the true posterior probability density of the target,and then use the sigma-like method to split and replicate the resampled particles.This method can effectively solve the particle diversity problem.The effectiveness of the proposed method in the detection and tracking of weak targets is verified by experimental simulation.